On-line monitoring device quality evaluation system based on fault tree and Bayesian network
Technical Field
The invention belongs to the field of quality evaluation and monitoring of online monitoring devices, and particularly relates to an online monitoring device quality evaluation system based on a fault tree and a Bayesian network.
Background
Through the development of the last decade, the on-line monitoring technology of the equipment is widely applied to the monitoring of the power equipment of the power grid, and plays an important role in ensuring the safety and stable operation of the power grid. The monitoring device can cover all or more than 220 kV and part of important 110 kV transformer substations and overhead transmission lines, can monitor key parameters of a main equipment body of the power transmission and transformation and key parameters of an external environment in real time, and supports technical personnel and management personnel of each level of operation and inspection to master the key parameters and states of the equipment; through monitoring unusual defective devices in real time, the operation and inspection personnel can be guided to timely overhaul the devices, the safe and stable operation of the power grid devices is guaranteed, the power failure times of the devices are reduced, the reliability of the devices is improved, and the device fault tripping is effectively reduced.
However, the online monitoring device is affected by external factors such as a complex strong electromagnetic environment of high-voltage equipment, a severe operating condition, lightning and the like for a long time in the operating process, and problems such as sensor abnormality, communication interruption, power failure, software defects and the like easily occur, so that data distortion and a high false alarm rate frequently occur, the captured equipment state abnormality cannot be captured in time, and the device itself is abnormal but alarms frequently. Under the condition, the light person causes false alarm and function failure of the device, and the field operation and maintenance difficulty is increased; the serious person causes misoperation of equipment protection and even causes safety production accidents. The main reasons for the problems are two aspects, on one hand, the product quality of part of manufacturers is inherently insufficient, so that various software and hardware faults are frequent; on the other hand, the device is lack of necessary operation and maintenance management means for monitoring the device, and the failure of the device is frequently caused due to the insufficient inspection and maintenance.
Therefore, the effectiveness and the reliability of the monitoring device are evaluated, the quality of online monitoring data is improved, and on one hand, operation and maintenance personnel can be helped to maintain and overhaul the online monitoring device in a targeted manner; on the other hand, the fault judgment opportunity of the power grid body can be effectively avoided from being delayed, and the running reliability of the power grid equipment is improved.
Therefore, an online monitoring device quality evaluation system based on a fault tree and a Bayesian network is provided.
Disclosure of Invention
The invention mainly aims to provide an online monitoring device quality evaluation system based on a fault tree and a Bayesian network, which can effectively solve the problems in the background art.
In order to achieve the purpose, the invention adopts the technical scheme that: the online monitoring device quality evaluation system based on the fault tree and the Bayesian network comprises the following steps:
step 1: selecting a sliding window according to the acquisition frequency and characteristics of the monitored quantity, and intercepting a data set of the forward time T from the current moment from the online monitoring data stream;
step 2: detecting data by repeating, missing, jumping and the like, if an abnormal mode exists, performing the next step, otherwise, returning to the step 1;
and step 3: judging a data flow abnormal mode and a situation that multiple kinds of abnormalities coexist possibly according to the device data abnormal model;
and 4, step 4: calculating a device fault Bayesian network by using the device fault statistical data as prior probability;
and 5: correcting a Bayesian network calculation result by using the data abnormal mode as supplementary information;
step 6: respectively calculating the ratio of the data volume of each abnormal mode to the normal data volume in the sliding window as a fuzzy index score;
and 7: establishing a fuzzy evaluation matrix according to the fuzzy index score in the step 6 and the fault probability of each component in the step 5;
and 8: and performing matrix multiplication operation on the weight set and the fuzzy evaluation matrix, and calculating the operation reliability evaluation result of the online monitoring device.
The construction of a preferred bayesian network comprises the steps of:
s1: establishing a device operation reliability evaluation index system, and defining an index set S ═ { S1, S2, S3, … …, S17} and a comment set V ═ V1, V2, V3}, wherein V1-V3 respectively represent normal, abnormal and fault states of the device;
s2: the index weight is calculated by taking into account the correlation between the index weight and the failure probability (P) and the degree of importance (I) of each index, and the item takes the index risk (R) as the basis for weight calculation, where R is P × I. The fault probability of each index can be obtained according to the calculation method in the previous section, the importance degree of each index can be calculated by using an analytic hierarchy process, and an index comparison matrix A is constructed by pairwise comparison of indexes A1-A9; secondly, calculating the characteristic vector and the relative weight of the matrix A; finally, consistency check is carried out on the relative weights, the consistency of results of the weights in pairwise comparison is guaranteed, the calculation results of the index weights are multiplied by the fault probability, and the results are normalized to obtain an index risk weight set omega which is { omega 1, omega 2, omega 3, omega 4, omega 5, omega 6, omega 7, omega 8, omega 9, omega 10, omega 11 and omega 12 };
s3: determining a membership function, wherein the item selects intermediate trapezoidal distribution, and the expression is as follows:
s4: establishing a fuzzy evaluation matrix, substituting all index scores into the formula to obtain the fuzzy evaluation matrix, namely
In the formula: rij represents the membership degree of the ith index to the jth comment; n represents the number of indexes; m represents the number of comments;
s5: calculating the operation quality evaluation result of the device, and performing matrix multiplication operation on the weight set omega and the fuzzy evaluation matrix R to obtain a model evaluation result, namely
And finally, according to the maximum membership criterion, the comment corresponding to the maximum value in the evaluation result B is the evaluation result.
Preferably, the A1-A9 indexes are pairwise compared to construct an index comparison matrix A, wherein A1: data interruption (instantaneous), wherein the abnormality is data loss, the duration is generally in the order of minutes, and the main reasons are software defects and poor communication quality;
a2: data interruption (long-term), compared with instantaneous interruption, the duration of the abnormality is generally of a daily level and mainly caused by communication interruption, battery electric energy exhaustion, software defects or parameter errors and the like;
a3: data repetition, wherein if the abnormality is caused by communication interruption, the data is generally a negative value or a limit value; if the sensor completely fails, the value is a random value in a certain measuring range. Such anomalies are typically not zero;
a4: data fixed offset, namely a fixed deviation exists between a detection value and an actual measurement value, and is caused by the fixed offset of the sensor;
a5: data is zero, and when the sensor sensitivity is reduced, a small change in the monitored quantity is difficult to capture, so the data remains zero;
a6: the data continuously increases/decreases, the data continuously increases/decreases along with time, and if the trend keeps the same direction for a long time, the trend is due to the drift of the sensor; otherwise, the device state is changed;
a7: data jumping, the equipment state changes in a step manner to cause the monitoring value to increase or decrease suddenly, and the monitoring value has certain continuity;
a8: data jitter, under severe operating conditions such as switch on/off and full load, the strong electromagnetic field which changes violently can cause data to fluctuate up and down with the mean value as the center;
a9: outliers, also known as discrete values, refer to values where the data deviates far from the statistical mean, and external environmental interference is a major source of such anomalies.
Preferably, the fault tree-based monitoring method includes the following steps:
step one, setting a time sliding window for real-time data acquired by an online monitoring device, and taking the data in the sliding window as a characteristic data set for evaluating the monitoring device;
step two, taking the distribution situation of abnormal values, the distribution situation of continuous identical values, the null value situation and the change situation of the variation coefficient in the feature data set as criteria, and applying the criteria to the feature data set;
acquiring corresponding discrimination values, and giving corresponding weight to each discrimination value according to the condition of the emphasis on the criterion;
and step four, adding and summing the discrimination values with the weights to obtain the state value of the monitoring device.
Compared with the prior art, the invention has the following beneficial effects: according to the online monitoring device quality evaluation system based on the fault tree and the Bayesian network, the field monitoring device does not need to be disassembled and inspected, and the normal operation of the power grid is not influenced; an online monitoring device state evaluation system is established, the device state is automatically evaluated, the investment of detection personnel is reduced, and the workload of first-line workers is reduced; the requirement on the experience level of workers is not high; the system analyzes the data acquired by the monitoring device, the data demand source is single, and the device diagnosis mode is simple and efficient.
Drawings
FIG. 1 is a model diagram of an evaluation index system for an online monitoring device constructed by FTA according to the present invention;
FIG. 2 is an evaluation index system model diagram of an online monitoring device constructed based on FTA.
FIG. 3 is a diagram of a Bayesian network constructed in accordance with the present invention.
Fig. 4 is a diagram illustrating the reliability evaluation procedure of the on-line monitoring device for power transmission and transformation equipment according to the present invention.
Detailed Description
In order to make the technical means, the creation characteristics, the achievement purposes and the effects of the invention easy to understand, the invention is further described with the specific embodiments.
As shown in fig. 1-4, the technical scheme adopted by the invention is as follows: the quality evaluation system of the online monitoring device based on the fault tree and the Bayesian network comprises the following steps:
step 1: selecting a sliding window according to the acquisition frequency and characteristics of the monitored quantity, and intercepting a data set of the forward time T from the current moment from the online monitoring data stream;
step 2: detecting data by repeating, missing, jumping and the like, if an abnormal mode exists, performing the next step, otherwise, returning to the step 1;
and step 3: judging a data flow abnormal mode and a situation that multiple kinds of abnormalities coexist possibly according to the device data abnormal model;
and 4, step 4: calculating a device fault Bayesian network by using the device fault statistical data as prior probability;
and 5: correcting a Bayesian network calculation result by using the data abnormal mode as supplementary information;
step 6: respectively calculating the ratio of the data volume of each abnormal mode to the normal data volume in the sliding window as a fuzzy index score;
and 7: establishing a fuzzy evaluation matrix according to the fuzzy index score in the step 6 and the fault probability of each component in the step 5;
and 8: and performing matrix multiplication operation on the weight set and the fuzzy evaluation matrix, and calculating the operation reliability evaluation result of the online monitoring device.
The construction of a preferred bayesian network comprises the steps of:
s1: establishing a device operation reliability evaluation index system, and defining an index set S ═ { S1, S2, S3, … …, S17} and a comment set V ═ V1, V2, V3}, wherein V1-V3 respectively represent normal, abnormal and fault states of the device;
s2: the index weight is calculated by taking into account the correlation between the index weight and the failure probability (P) and the degree of importance (I) of each index, and the item takes the index risk (R) as the basis for weight calculation, where R is P × I. The fault probability of each index can be obtained according to the calculation method in the previous section, the importance degree of each index can be calculated by using an analytic hierarchy process, and an index comparison matrix A is constructed by pairwise comparison of indexes A1-A9; secondly, calculating the characteristic vector and the relative weight of the matrix A; finally, consistency check is carried out on the relative weights, the consistency of results of the weights in pairwise comparison is guaranteed, the calculation results of the index weights are multiplied by the fault probability, and the results are normalized to obtain an index risk weight set omega which is { omega 1, omega 2, omega 3, omega 4, omega 5, omega 6, omega 7, omega 8, omega 9, omega 10, omega 11 and omega 12 };
s3: determining a membership function, wherein the item selects intermediate trapezoidal distribution, and the expression is as follows:
s4: establishing a fuzzy evaluation matrix, substituting all index scores into the formula to obtain the fuzzy evaluation matrix, namely
In the formula: rij represents the membership degree of the ith index to the jth comment; n represents the number of indexes; m represents the number of comments;
s5: calculating the operation quality evaluation result of the device, and performing matrix multiplication operation on the weight set omega and the fuzzy evaluation matrix R to obtain a model evaluation result, namely
And finally, according to the maximum membership criterion, the comment corresponding to the maximum value in the evaluation result B is the evaluation result.
Preferably, the A1-A9 indexes are pairwise compared to construct an index comparison matrix A, wherein A1: data interruption (instantaneous), wherein the abnormality is data loss, the duration is generally in the order of minutes, and the main reasons are software defects and poor communication quality;
a2: data interruption (long-term), compared with instantaneous interruption, the duration of the abnormality is generally of a daily level and mainly caused by communication interruption, battery electric energy exhaustion, software defects or parameter errors and the like;
a3: data repetition, wherein if the abnormality is caused by communication interruption, the data is generally a negative value or a limit value; if the sensor completely fails, the value is a random value in a certain measuring range. Such anomalies are typically not zero;
a4: data fixed offset, namely a fixed deviation exists between a detection value and an actual measurement value, and is caused by the fixed offset of the sensor;
a5: data is zero, and when the sensor sensitivity is reduced, a small change in the monitored quantity is difficult to capture, so the data remains zero;
a6: the data continuously increases/decreases, the data continuously increases/decreases along with time, and if the trend keeps the same direction for a long time, the trend is due to the drift of the sensor; otherwise, the device state is changed;
a7: data jumping, the equipment state changes in a step manner to cause the monitoring value to increase or decrease suddenly, and the monitoring value has certain continuity;
a8: data jitter, under severe operating conditions such as switch on/off and full load, the strong electromagnetic field which changes violently can cause data to fluctuate up and down with the mean value as the center;
a9: outliers, also known as discrete values, refer to values where the data deviates far from the statistical mean, and external environmental interference is a major source of such anomalies.
Preferably, the fault tree-based monitoring method includes the following steps:
step one, setting a time sliding window for real-time data acquired by an online monitoring device, and taking the data in the sliding window as a characteristic data set for evaluating the monitoring device;
step two, taking the distribution situation of abnormal values, the distribution situation of continuous identical values, the null value situation and the change situation of the variation coefficient in the feature data set as criteria, and applying the criteria to the feature data set;
acquiring corresponding discrimination values, and giving corresponding weight to each discrimination value according to the condition of the emphasis on the criterion;
and step four, adding and summing the discrimination values with the weights to obtain the state value of the monitoring device.
The foregoing shows and describes the general principles and broad features of the present invention and advantages thereof. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.